Convergence rates for probabilities of moderate deviations for moving average processes (Q943502)

Let \(\{Z_n:n\geq1\}\) be an arbitrary sequence of real random variables and let \(a_n,b_n\) be positive constants. The authors prove that complete moment convergence in the sense of \textit{Y. S. Chow} [Bull. Inst. Math., Acad. Sin. 16, No. 3, 177--201 (1988; Zbl 0655.60028)] \[ \sum_{n=1}^\infty a_n E[b_n^{-1}|Z_n|-\varepsilon]_+^q <\infty \qquad \forall\varepsilon>\varepsilon_0\geq0 \] for some \(q>0\) is equivalent to a refined version of complete convergence by \textit{D. Li} and \textit{A. Spătaru} [J. Theor. Probab. 18, No. 4, 933--947 (2005; Zbl 1085.60013)] \[ \int_\varepsilon^\infty\sum_{n=1}^\infty a_n P\{|Z_n|>x^{1/q}b_n\}\,dx<\infty \qquad \forall\varepsilon>\varepsilon_0^q\geq0. \] In fact a more general version of the equivalence is shown. In case of partial sum processes \(Z_n=\sum_{i=1}^nX_i\) with certain dependence structure among the random variables \(\{X_n:n\geq1\}\) by a comparison of known results on complete moment convergence and (refined) complete convergence, both remain valid under the same conditions. Remarkably, no assumption on the dependence structure is needed here. The authors further give moderate deviation results in terms of (refined) complete convergence (and thus on complete moment convergence) under various moment conditions for partial sums of moving average processes \(X_n=\sum_{i=-\infty}^\infty a_{i+n}Y_i\), where \(\{Y_i:i\in\mathbb Z\}\) is an i.i.d. sequence and \(\{a_i:i\in\mathbb Z\}\) is an absolutely summable sequence.